Google says its new Nest Wi-Fi isn’t just Google Wi-Fi plus a smart speaker—it’s new, improved, and with better coverage. This is good news—despite the impressive sales numbers Google says it has for the original Google Wi-Fi, the product never ranked very well in performance tests at Wirecutter, Smallnetbuilder, or here at Ars.
In fact, Google claimed that the new Nest Wi-Fi would cover “up to 3,800 square feet.” This raised our eyebrows immediately, since one square foot is very much not like another—my own 3,500-square-foot house is a particular bear to cover due to the downstairs floor having plenty of packed earth occluding the line-of-sight between that entire floor and the router, for example.
If you set up Wi-Fi systems in many houses, you’ll also soon discover that even without an occluded sub-basement floor, a house shaped like one long Viking hall is very different from a house with similar volume but arranged in a mostly rectangular or L-shape. And heaven help you once you get into turn-of-the-century chicken-wire-and-daub walls, or apartments with central elevator shafts, and so forth.
The good news is, a two-piece Google Wi-Fi system did cover even my challenging 3,500-square-foot house—an impressive achievement and a clear win over the original Google Wi-Fi.
Nest Wi-Fi is all new
It wasn’t clear from the launch announcements, but Nest Wi-Fi really is more than just Google Wi-Fi plus a smart speaker. The Nest Router is a 4×4 AC2200 access point and router, and the Nest Point is a 2×2 AC1200 access point. Each of the original Google Wi-Fi nodes was also an AC1200 2×2 access point, but Nest Wi-Fi points are updated with a newer and better wireless chipset, with additional RAM and a CPU nearly twice as fast as the original’s. Granted, a lot of that new CPU power and RAM is really there to power the Google Assistant—but also it comes in handy for general network use.
Google Nest Wi-Fi
The original Google Wi-Fi really needed at least three nodes to adequately cover the test house, but Nest Wi-Fi managed it credibly with two. Google does still sell three-piece kits, as well as individual Points and Routers that can be added into existing networks, and we hope to update and retest with an additional Nest Point soon.
Although Nest Wi-Fi isn’t just Google Wi-Fi with a smart speaker, the Assistant is a big piece of Google’s sales pitch for the new kit. We don’t feel there’s much to say about it that hasn’t already been said about the other Google Home Assistant products, though.
If you own a Google Home smart speaker, the Assistant in Nest Wi-Fi is going to seem very familiar—in everything from overall sound to listening ability to some occasional bugginess. The smart speaker functionality isn’t in the Nest Router, though—only in the Nest Points. It’s loud without being overpowering, sounds decent without really wowing you (think “that cheap jam box you had in college and liked okay”), and it listens well and accurately from significant distance. It seems to have a hard time coping with hearing around physical obstructions, though.
For the most part, the speaker in our Nest Point worked as you’d expect—we set timers, asked it to tell us jokes, and other normal things. With the Point atop the ten-foot tall island in the living room, it pretty much refused to accept voice commands unless we performed a chin-up and spoke to it without the cabinet in the way. We don’t hold that against the device, though; while atop the island is the best place in my house for the second node in a two-node mesh kit, we don’t expect many people will have similar home layout issues.
Sometimes, though, the Assistant was just plain stubborn. When asked “Hey Google, play House of Pain,” our assistant said “Sure! Playing House of Pain on Google Play Music”—and then didn’t play anything. After five tries in a row, we tried “Hey Google, play ‘Jump Around’ by House of Pain,” and that worked. After that, asking for just the band and not a particular song worked fine.
We don’t think this has anything to do with the Nest Point specifically, though, since sometimes the Google Assistant on Pixel phones gets similarly cranky.
Prepare for a long setup process
One of the drawbacks of bundling a Google Assistant with a mesh Wi-Fi kit is the sheer length of the setup process necessary to get both going at once. We were already familiar with the setup process for both Google Home and Google Wi-Fi but still ended up needing better than half an hour to get everything operational. We took thirty setup screenshots along the way—and deleted an awful lot more. There’s a lot of “next” and “okay” and deciding whether to accept multiple different kinds of telemetry to get through before you can start playing.
If you’re really, really excited about your new Nest Wi-Fi kit, this might be okay. For us, it produced significant “setup fatigue”—we feel this process could be streamlined substantially.
We’ll spare you the entire panoply of steps, because there’s nothing particularly interesting or difficult about them—it’s just a lot of steps, a lot of time, and a lot of different privacy asks that seem like they really could have been combined into one section. We did note that Nest claims its voice print matching technology has gotten better—it noted that it had a voice print on file for me already, due to my playing around with a Google Home speaker in my office—but said it had “upgraded” the voice print and it would work “better” now. This did not necessitate any additional voice recordings.
Testing the Wi-Fi
At Ars, we’re always on the lookout for ways to improve our testing metrics. In the past, we’ve used netburn to variously model 4K streaming, HTTP downloads, Web browsing, and more. Each workload stresses a Wi-Fi network in different ways—but if your tests get too complex, they’re difficult to interpret and may also not test all sites in your environment equally.
We think we’ve improved things significantly this time around by cutting it down to four basic tests.
- Full rate download, all stations (STAs) simultaneously—in this test, we repeatedly download a 1MB file via HTTP as fast as possible from each of our four test stations (laptops) at once. Ideally we’ll not only see lots of throughput, we’ll see relatively equal throughput at all sites. Some Wi-Fi routers or mesh kits will allow one STA to “starve” others when they’re all clamoring for bandwidth at once, and that’s not what we want to see here.
- Full rate download, each STA solo—in this test, we repeatedly download a 1MB file via HTTP at each STA. Each station gets to fly solo, though—this is what you can expect for simple throughput in an uncongested network where you’re not competing with neighbors, roommates, spouses, IoT gear, etc. for airtime.
- 1080P streaming and browsing torture test—this is where it starts getting interesting! In this test, all four STAs emulate both a 1080P streaming session (fetching 1MB chunks of data by HTTP, rate-limited to 5Mbps by sleeping between fetches as necessary) and also a Web browsing workload (fetching “webpages” at 1Mbps; where each “page” consists of sixteen 128KB files, fetched in parallel over HTTP). We monitor both the streaming and browsing results. For the streams to succeed, they need to average 4.5Mbps or better; for the browsing to succeed, the mean latency should be under 1000ms maximum.
- 4K streaming and browsing torture test—this is just like the 1080P torture test, but we’re requesting 20Mbps per stream instead of 5Mbps. Few Wi-Fi networks will be able to feed four separate 4K streams to distant stations simultaneously. Even fewer will be able to do so and provide good Web browsing latency while they do.
Placement of STAs and APs
Our four test laptops are located:
- Along the inside wall of the large upstairs bedroom
- Underneath the living room TV (where a set-top box might go)
- In the lower-right corner of the kitchen
- Along the outside wall of the downstairs master bedroom
The downstairs master bedroom is a particularly obnoxious place to cover, since it’s blocked from the router by fifty feet or more of packed earth and a concrete slab. Even the second node of a two-piece kit has to reach it by way of both a wall and a floor/ceiling at a nasty oblique angle.
The upstairs master bedroom is also frequently a source of pain, as it’s quite distant from both the router and second node. Multiple interior walls, cabinets, and more must be traversed to reach the upstairs master from either node.
The APs—access points, as in the Nest Router and Nest Point—are placed in the router closet off the entryway hall and atop the big island in the living room, above the TV. The spot atop the island in the living room is by far the best place for a second point, since it can both reach the router closet with minimal distance and with a pretty clean line of sight, and also cover the downstairs floor directly.
That does it for Nest Wi-Fi—but since we updated our test protocols, we also needed a guinea pig to compare Nest Wi-Fi to. Since Plume’s Superpods are our all-time heavyweight champions, we dusted them off and set them up again as well. Our Plume gear is a four-piece kit, and we placed one Superpod in the router closet, another atop the TV island, a third on the inside wall in the upstairs master bedroom, and a fourth in the den downstairs, directly beneath the living room TV.
We were running short on time, so Plume did not get its usual 24-hour break-in and training period—we just blew the dust off, plugged it in, and started testing right away.
Cage match—Nest Wi-Fi vs Plume Superpods
In one sense, this is an incredibly unfair contest—we didn’t even get the three-piece Nest Wi-Fi kit, so we’re pitting a $ 270 two-piece kit vs a $ 400 four-piece kit. That’s without even getting into Plume’s annually billed membership, which does offer a lot of features but adds to the cost. The whole “pay a subscription for your Wi-Fi” also rankles a lot of tech folks.
On the other hand, Google claims its two-piece kit would cover up to 3,800 square feet—so we gave it the chance to put its Wi-Fi where its mouth is.
When looking at the absolute simplest metric—how much raw throughput can you get at each laptop when that STA is the only one active—Nest Wi-Fi hangs reasonably well with the Superpods. It did struggle in the downstairs master bedroom, but it was at least in the same ballpark with the Superpods everywhere else—it even beat them in the kitchen.
The real story when it comes to Wi-Fi isn’t single device throughput, though—the real pain happens when you’re competing with other devices. When we asked for all the throughput we could get from all four stations simultaneously, things got hairier. In the simultaneous throughput test, Nest Wi-Fi provided considerably less throughput in all locations than Plume did.
We still believe the real test of a Wi-Fi network isn’t throughput at all—it’s browsing latency, especially when the network is already congested. To that end, we ran simultaneous 1080P emulated streams and 1Mbps emulated webpage fetches on all four stations simultaneously. Nest Wi-Fi mostly succeeded here in providing all four 1080P streams, though it’s still struggling with that downstairs bedroom. However, things go off the rails when we look at browsing latency.
It took nearly four seconds on average for Nest Wi-Fi to deliver a webpage to the downstairs bedroom, and about 1500ms to deliver one to the kitchen. It’s worth noting that this really is an average—meaning many of the fetches took longer than that. That’s not a good experience. And while we call this a “torture test,” it’s really not that far off from a normal day at many residences. There are two adults and three kids in my house, and it’s not that uncommon for at least four of us to be watching different things on Rokus, tablets, and phones at once, while another of us tries to browse social media.
Making matters worse, both of the Rokus at my house will default to 4K where available—and more and more of Netflix’s catalog is 4K. So for the last pair of tests, we stepped things up and asked for four (emulated) 4K streams—requiring 20Mbps apiece—and, again, tried to do some (emulated) light Web browsing at the same time. This was a Herculean task because two of our four test stations were deliberately placed in very difficult-to-reach areas, and we didn’t really expect either kit to succeed.
Nest Wi-Fi was able to deliver simultaneous 4K streams to the living room and the upstairs bedroom, and it probably succeeded in the kitchen as well. Although the kitchen didn’t quite get the 20Mbps it asked for, anything more than 15Mbps generally suffices for real-world streaming.
Once again, though, Nest failed in the downstairs bedroom. At 6.5Mbps, it is a third of what we requested and less than half the bare minimum necessary for successful 4K streaming. It was no surprise that Nest Wi-Fi didn’t do well on browsing latency during the 4K torture test, either—again, we expect that nothing short of wired LAN or at least wired APs could handle this workload.
But to our surprise, the Plume Superpods succeeded on both halves of the 4K torture test, delivering 19.9Mbps to each STA and supplying webpages with under 500ms browsing latency in all four locations.
Nest Wi-Fi is a big improvement over the original Google Wi-Fi—and, true to its claims, it did cover our unusually difficult 3,500-square-foot test house. Simply “covering” a house isn’t the same thing as making everything work the way you want it to, though. We think many home users will expect a mix of video streaming and Web browsing by everyone in the house at once to work seamlessly.
Nest Wi-Fi is capable of getting that job done under most circumstances, but there’s still room for improvement with how quickly, consistently, and reliably it does so in challenging environments. If all you’re looking for is great Wi-Fi, you can do better—we don’t see the value in buying Nest Wi-Fi if you aren’t completely sold on the integrated smart speaker.
Given the price point and the good-but-not-amazing Wi-Fi in Nest, we see it as best suited to those people who are all-in on voice commands. If you want smart speakers everywhere and good Wi-Fi coverage, it’s going to be difficult to beat Nest as an all-in-one product.